Value for an Array that will not match any other Array.

[paulmansour]

So I want to initialize a variable that cannot match any other value provided by a user (another programmer, or anything really).

I came up with an empty namespace:

      ⎕NS ''

I think it is impossible to proved a value that will match the result of this.

I think there is no other value A(other than a class instance of some sort).

Am I right in both cases?

[StefanoLanzavecchia]

I asked myself the same question a few years ago and I came out with the same answer: an anonymous namespace is a wonderful unique value. It's a scalar and it can be given a reasonable appearance with []DF.
Its presence anywhere obviously blows the size of any until then homogenous array, but if used in in well controlled contexts is just marvelous.
I never had the problem of persisting it (in DCF files or send it through sockets) so I have no idea how it behaves there, if its identity it's preserved and so on.

[AndyS|Dyalog]

I have just added a check to the QAs to ensure that

      (⎕ns'')≢⎕ns''
1

with a comment that changing this would cause you problems.

The ideal solution should perhaps be based on associating UUIDs/GUIDs with the variables, but given that we do not have a built-in cross-platform mechanism for generating them, you would have to make a ⎕NA call or two to get them.

[Phil Last]

I too have used this mechanism but I doubt if

      z={N+←1 ⋄ ⎕ns''}⍣{⍺=z}z←⎕ns''⊣N←0

would not eventually return a 1, with N being ¯1+2*32 or some such.

[AndyS|Dyalog]

I asked the developers about this, and they replied that each time you call ⎕ns'' the interpreter creates a new, independent object in the workspace. When comparing two refs, the interpreter looks to see if they are pointing to the same object in the workspace. Since they're not, the comparison fails.

I have been running your expression since yesterday; when I interrupted it N had a value greater than 2*33 ..

[JoHo]

Guessing about the implementation of ⎕NS in the background,
this could be running on only a few pooled unnamed namespaces:
could it be, that un-referenced NS is either re-used at the next loop or
garbage collected.
Would it make a difference to run this test like this:

Code: Select all

 z= {N+←1 ⋄ NSL,←⎕NS''}⍣{⍺=z}z←⎕NS''⊣ N←0⊣ NSL←''

EDIT1: corrected code to

Code: Select all

⍣{⍺=z}

[Phil Last]

Would it make a difference to run this test like this:

I too wondered about gathering them but I guess WS FULL would ensue first.

when I interrupted it N had a value greater than 2*33 ..

Me too!

creates a new, independent object in the workspace

I guess that settles it. Good!

      )clear
clear ws
      )wsid ws
was CLEAR WS
      this←⎕ns''
      )save
ws saved Fri Nov 27 13:39:53 2015
      that←this
      )copy ws this
.\ws saved Fri Nov 27 13:39:53 2015
      this=that
0
      ⊢'cf'⎕fcreate 0
1
      this ⎕fappend 1
      that←⎕fread 1 1
      this=that
0

Fairly convincing.